Research Of Color Image Retrieval Based On Color And Shape

Along with the arrival of multimedia era, people can get more and more image information. Therefore, how to provide an efficient approach to query these images which have rich content becomes an active hot point in the area of retrieval research. Content based image retrieval (CBIR) is just one of key technologies for such a problem.Visual feature extraction is the basis of CBIR technology. The visual feature includes color, shape and texture characteristics. In this paper, the color and shape feature extraction has been paid much attention. Typically, color histogram is used to describe the color feature. But in order to obtain the color histogram of an image, color quantization should be executed first. Thus, in this paper, a new algorithm for color quantization is proposed. The main idea of the algorithm is to get the palette from sample image by unsupervised color quantization, and then get color feature of images in the image library by supervised color quantization. Analysis and experiments have proved that the algorithm has a lower false ratio and faster rate. However, only using color feature can not describe the spatial distribution of the image color. So how to describe it has been studied primarily and a new shape description method is put forward. The method uses the number of pixels of the color on the border to describe its intensity and can obtain satisfied results. In order to obtain better retrieval results, the color features are combined with the shape characteristics to describe image content in this paper. The experiment results have shown that this method has a very good retrieval results.